Agentic GraphRAG: Multi-Agent Knowledge Graph
summary
A reproducible multi-step system ingests PDFs and tables, runs a four-agent pipeline (analyzer, extractor, merger, optimizer) with OpenAI Chat Kit and LangSmith tracing, and persists cleaned graphs to Neo4j. From five 8k-character chunks it extracts ~84 entities then semantically merges to 32, writes 57 nodes and 179 relationships in ~2s using batch UNWIND+MERGE, and flags/removes low-confidence relations with an LLM QA pass for higher precision.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:48Current challenges
- 3:57Architecture overview
- 6:15User input
- 7:56Research analyzer agent
- 10:34LangSmith tracing
- 12:15Extractor agent
- 14:46Entity merger agent
- 17:04Graph optimizer agent
- 19:01Neo4j storage
- 24:50Future improvements
related talks
12mAgentic GraphRAG: Autonomous Knowledge Graph Construction and Adaptive Retrieval
21mAgentic GraphRAG: Autonomous Knowledge Graph Construction and Adaptive Retrieval
30mFusing NLP and Graph: Building a Conversational Agent for Enriched Financial Data
01h 26mThinking in Graphs: The Cognitive Foundations of Agentic AI
36mGraphRAG Explained 🔥 Building Smarter LLMs
30mHow to Build the Next Wave of Graph-Based Agentic AI
53mPractical GraphRAG - Making LLMs smarter with Knowledge Graphs
20mPractical GraphRAG: Making LLMs smarter with Knowledge Graphs